David A. C. Beck

This study integrates sparse autoencoders into ProteinMPNN to uncover protein features and improve thermostability of a variant of Brolucizumab.

This paper introduces NOMELT, a neural machine translation model that designs high-temperature stable proteins by learning thermophilicity from protein homologous pairs.

The paper curates a large dataset of protein-temperature pairs and demonstrates its potential for deep learning to improve thermal stability design.